Why Emails Go to Spam

The layered model

Major receivers like Gmail, Outlook, and Yahoo apply spam decisions in roughly this order. Each layer can either drop the message or pass it on with a score modifier. Failures in earlier layers carry far more weight than later ones.

Layer 1: Connection-level failures

Sending IP on a blocklist

The most binary failure mode. If your sending IP is listed on a major blocklist (Spamhaus SBL/XBL, SORBS, Barracuda, etc.), connections are refused at SMTP. Mail never reaches the recipient mailbox.

To check: paste your IP into MXToolbox Blacklist Check or MultiRBL.valli.org. Either tool queries 50-100 blocklists simultaneously. Most listings provide a delisting request form. Spamhaus listings often require fixing the underlying reason (compromised server, open relay, spam complaints) before delisting is granted.

Missing or bad reverse DNS (PTR)

Every IP that sends mail should have a PTR record (reverse DNS) that resolves to a hostname which itself resolves back to that IP (FCrDNS — forward-confirmed reverse DNS). Without this, many receivers refuse the connection or score the message heavily. The PTR is set by whoever controls the IP — typically your hosting provider, not you. For cloud mailservers (Google, SendGrid, etc.) the provider handles this automatically.

No or weak TLS

Connections without STARTTLS are scored as suspicious by most receivers in 2026. Connections with weak ciphers (SSL 3.0, TLS 1.0/1.1) or expired/self-signed certificates may be accepted but reduce sender reputation over time. Modern mailservers should support TLS 1.2 or 1.3 with valid certificates.

Layer 2: Authentication failures

See the dedicated guides for full detail:

• SPF failures — IP not authorized, 10-lookup limit exceeded, multiple SPF records.
• DKIM failures — signature does not verify, key rotation issue, body modification in transit.
• DMARC failures — no alignment between authenticated domain and visible From: domain.

The Gmail/Yahoo 2024 requirements make this layer particularly important: bulk senders without aligned SPF + DKIM + DMARC are now actively filtered, not just scored down.

Layer 3: Sender reputation

This is the single most common reason that previously-delivering mail suddenly starts going to spam. Reputation tracks how recipients have reacted to your mail over a rolling 30-90 day window.

Complaint rate

The most heavily weighted reputation signal. Every "Report Spam" button click counts against you. Gmail's threshold for healthy sending is below 0.1% spam complaint rate; above 0.3% is the Gmail/Yahoo 2024 noncompliance line. To monitor: enroll in Google Postmaster Tools and watch the spam rate graph daily.

Engagement rate

Modern receivers infer engagement from open rates, click rates, replies, and time-spent-with-message-open. Mail to people who never open it is treated as low-value even without complaints. The first 1000 messages of any new sending campaign are weighted heavily — receivers calibrate their reputation models on early traffic.

Volume changes

Sudden volume spikes from a previously low-volume domain look like a hijacked account or a poorly-warmed-up sender. Ramp gradually — doubling daily volume is roughly the maximum safe rate when starting from cold. Going from 100 messages/day to 10,000 messages/day in one campaign reliably triggers filtering.

List hygiene

Sending to bounced or unknown addresses raises your "unknown user" rate, which receivers treat as a strong signal of poor list hygiene — usually associated with purchased lists, which are correlated with spam. Hard bounces (recipient does not exist) must be removed from your list immediately. Soft bounces (mailbox full, temporary issue) can be retried a few times then removed if persistent.

Layer 4: Content signals

Content scoring in 2026 is dominated by machine-learning models rather than keyword lists. The signals that matter:

URL reputation

Every URL in the message is checked against URL reputation databases. URL shorteners (bit.ly, tinyurl) are treated as suspicious by default because they hide the destination. Brand-impersonating URLs (typo domains, look-alike domains) score highly as spam. Newly-registered domains and domains with no traffic history are suspicious. Best practice: link only to your established domain or known-good external sites.

Image-to-text ratio

Messages that are mostly images with little text trigger spam scores. Marketers sometimes embed text as images to bypass keyword filters; receivers recognize and penalize this. Aim for a healthy balance — significant body text in addition to any images.

Attachments

Executable attachments are blocked outright. Encrypted ZIPs are heavily suspicious. PDFs and Office documents are scanned for embedded scripts and known malicious patterns. Large attachments (>5MB) on transactional mail are unusual and contribute to spam scoring.

HTML quality

Malformed HTML, missing alt text on images, mismatched display text and link URLs ("click www.legitimate-bank.com" style), and obfuscated text (zero-width characters, mixed-script Unicode) all reduce reputation. Templates from major email service providers handle this correctly; hand-coded HTML is where problems usually start.

Subject line

Subject lines no longer trigger filtering on simple keywords like "free" or "win" — that was a 2010 problem, not a 2026 one. ML-based subject scoring looks at full semantic meaning and historical patterns. ALL CAPS, excessive punctuation, and emoji overload contribute weakly to spam scores but the larger issue is that they correlate with recipient complaint behavior. If recipients see and click "Report Spam" because of an aggressive subject, your reputation suffers regardless of the subject's content score.

Layer 5: Per-recipient signals

This is the layer most operators forget exists. Each individual mailbox has its own filtering state for each sender:

• If a recipient marked you as spam, future mail from your domain goes to their spam folder regardless of your global reputation.
• If a recipient consistently opens your mail and never complains, you get inbox placement even when your global reputation is borderline.
• If a recipient added you to their contacts or replied to a previous message, you get strong delivery preference.

This explains the common observation that the same campaign lands in inbox for some recipients and spam for others. Global reputation is the floor; per-recipient signals are the modifier.

The diagnostic workflow

1. Test with mail-tester.com. Send a test message to the unique address it provides. The tool returns a 0-10 score with a breakdown of every signal: authentication, blocklists, content, HTML quality, headers. Free for 3 tests per day per IP.
2. Enroll in Google Postmaster Tools. Required for any serious mail operator. Shows daily reputation, spam rate, authentication metrics, encryption rate, and IP reputation for your domain. Free.
3. Read your DMARC aggregate reports. Daily XML reports show every sender claiming to be you and their authentication results. Use a parser (dmarcian, EasyDMARC, parsedmarc).
4. Check Microsoft SNDS. Smart Network Data Services shows reputation data for IPs sending to Microsoft (Outlook.com, Hotmail). Less data than Google Postmaster Tools but essential if Microsoft is a significant chunk of your recipient base.
5. Read bounce DSNs carefully. Every bounce has a delivery status notification with the reject reason. Patterns in the DSN text (5.7.1 = policy reject, 5.1.1 = user does not exist) point at different fixes.

The recovery playbook

If you are already in spam-folder hell, here is the order that works:

1. Stop sending to inactive recipients. Cut your list to people who have engaged in the last 90 days. This is non-negotiable; sending to dead addresses keeps reputation low.
2. Fix authentication. Confirm SPF, DKIM, DMARC are all passing and aligned. Aggregate reports will tell you if anything is broken.
3. Resolve any blocklist listings. Submit delisting requests and fix root causes.
4. Re-warm your sender. Send only to engaged recipients for 2-4 weeks. Keep volume low and steady. Engagement signals must outweigh any historical complaints in the rolling window.
5. Audit content. Use mail-tester.com on representative messages. Address any issues flagged in the content section.
6. Wait. Reputation moves on rolling windows. There is no override; consistent good behavior over weeks is the only fix.

Frequently Asked Questions

Why does my email go to spam only for some recipients?

Each mailbox provider runs its own filter with its own reputation database. Gmail might trust your domain while Outlook does not, because they have seen different volumes of mail from you and different complaint rates. Per-recipient spam placement also depends on individual user behavior — if a specific recipient has historically marked mail from your domain as spam, future mail from you goes to their spam folder regardless of your reputation elsewhere.

Does using ALL CAPS in subject lines cause spam filtering?

ALL CAPS subjects do not directly trigger spam filtering in 2026 — Bayesian and ML-based filters have largely moved past keyword pattern matching. However, ALL CAPS does correlate with low-quality marketing mail in training data, so it contributes weakly to spam scores. The bigger issue is recipient behavior: many users see ALL CAPS subjects and report as spam, which raises your complaint rate — that genuinely hurts deliverability.

How do I check my sender reputation?

For Gmail: enroll in Google Postmaster Tools (postmaster.google.com), which shows daily domain reputation, IP reputation, spam rate, and authentication metrics. For Microsoft: SNDS (Smart Network Data Services) and JMRP for IP reputation. For overall: check blocklist tools like MXToolbox Blacklist Check, Sender Score (Validity), and Talos (Cisco). Postmaster Tools is by far the most useful — it shows trends so you can correlate reputation changes with sending behavior.

Does sending the same email to many recipients trigger spam filters?

Identical content sent to many recipients is not inherently spammy — that is how newsletters work. What triggers filters is the combination of identical content + low engagement (low open rate, high complaint rate). If recipients engage (open, click, reply), high-volume identical content delivers fine. If recipients ignore or report, filters learn to send future similar content to spam regardless of volume.

How long does it take to recover from being marked as spam?

Two to eight weeks of consistent good behavior. Stop sending to unengaged recipients, fix the underlying problem (authentication, content, complaint source), then send only to engaged users for several weeks. Reputation recovers gradually as recent sending history outweighs the bad period. Quick fixes do not exist — reputation is based on rolling windows of 30-90 days at most major receivers.